Managing shared features across multiple HA instances — is there a better approach than AppDaemon?

[royal bridge]

I run Home Assistant on several locations (different houses) and have common "features" I want to share across all of them. By feature I mean a bundle of: helpers/entities + automation logic, ideally with a config file to parameterize it per house (different entity names, thresholds, etc.).

My current setup uses AppDaemon: I write the logic in Python, parameterize everything via apps.yaml, and it's easy to debug and adapt. However it has friction points about the creation of entities and divergence with core features of ha.

I'm wondering: is there a more "native" or widely adopted way to do this in the HA ecosystem?

I've looked at:

• Packages — great for YAML bundles but can't handle stateful Python logic
• Custom components — more native but heavier to write and iterate on
• Blueprints — good for parameterized automations but limited scope
• PyScript — promising Python approach, but unclear on entity management and multi-instance deployment

Am I missing something? How do others manage feature reuse across multiple HA instances?

[patent gyro]

There is a custom integration Remote Home-Assistant that allows instances to be linked and share some stuff. I haven't used it personally and I am not totally sure on its features but it might be worth a look?

[royal bridge]

Thanks for the suggestion! Though I think it doesn't quite fit my use case — Remote Home-Assistant is about linking instances together at runtime (sharing entities/state between them), whereas my problem is more about deploying the same feature code to independent instances. The houses don't need to talk to each other, they just need to run the same logic.

[patent gyro]

Gotcha, So here is an idea... you could store your config files on github then use a shell_commaand to run git to clone the files down. The files will then be in a known place which you can include to the config. Then you could run the clone and a yaml refresh on a schedule with an automation (or on demand by some means).

I am not sure if it would work but perhaps its a place to start?

[polar dragon]

I think you could even set up a site specific substitutions file linked to the actual entities present and then keep everything else generic